This page explains, for a UK shop owner rather than a developer, what people actually mean when they say they run their online shop from ChatGPT or Claude. It covers what an assistant can read, what it can change, what you should never let it do unattended, and the steps to connect one to an Orbit shop. Every technical claim below was checked on 15 September 2026 against modelcontextprotocol.io, developers.openai.com and support.claude.com. We build Orbit Commerce, so we have a stake in you opening a shop; the sources are linked so you needn’t take our word for it.
The two things people mean by “AI running my shop”
Two different products hide behind that phrase. One faces your customers. The other faces you.
A chatbot for customers sits on your storefront, answers “where is my order” and hands the hard ones to you. The risk is that it says something wrong to a stranger.
An assistant that works your admin is the opposite. Nobody but you sees it. You open ChatGPT or Claude, type “which products sold best last month”, and it reads your shop and answers. Type “put the slow ones in a 15% sale” and, if you gave it permission, it does that too. It replaces the tabs you click, not the emails you send. This post is about the second one.
What is MCP, in one sentence?
MCP (the Model Context Protocol) is, in its own words, “an open-source standard for connecting AI applications to external systems” (modelcontextprotocol.io, checked 15 September 2026). That is all it is: an agreed way for an assistant to plug into something else, the way USB-C is an agreed way to plug in a charger. The same page lists Claude and ChatGPT among the applications that support it.
Why it matters to you: because it’s a standard rather than a private deal, your platform publishes one connection and any assistant that speaks MCP can use it. You are not buying an “AI feature”. You are pointing the assistant you already pay for at your own shop.
What can an AI assistant actually do to your shop: read, or write?
Both, but only what you tick. In Orbit the permissions come in three tiers: read, safe write and live write. Read means it can look and report. Safe write means recoverable changes, like a draft product or a description. Live write means things a customer will feel, like a price change or a refund. The tiers exist so you can say yes to the first two and no to the third. Here is the same idea applied to jobs people actually ask for:
| What you ask for | Read or write | How we’d handle it |
|---|---|---|
| “Which products sold best last month?” | Read | Safe to leave on permanently. Nothing changes. |
| “Add these 40 products from a spreadsheet” | Write | Let it create them as drafts, read them yourself, then publish. |
| “Change the price of everything in the sale collection” | Write, live | Ask for the list of what it will change first. Approve, then run. |
| “Refund order 1042” | Write, live and irreversible | We wouldn’t grant this. Money leaving is a human decision. |
| “Rewrite every product description in a friendlier tone” | Write, bulk | Do ten, read them, then do the rest. Keep the old copy exported. |
Writes vary enormously by blast radius: a draft product affects nobody, a bulk price change hits every shopper within seconds, and a refund moves money you cannot pull back.
What should you never let an assistant do without approving it first?
Our line: never grant unattended permission for anything that moves money, touches a customer’s personal data, or changes hundreds of records in one go.
- Refunds and payouts. An assistant misreading “refund the damaged one” as “refund them all” is a bank transfer, not a typo.
- Cancelling or editing paid orders. Same reason, plus a customer at the other end.
- Bulk price changes without a preview. Ask what it intends to change and read the list. Every serious bulk edit should be a two-step with you in the middle.
- Deleting anything. Archive instead.
- Emailing your customers. Drafting is fine. Sending is you.
Everything else is fair game, and that “everything else” is most of your week: reports, stock questions, product creation, descriptions, collections, redirects and feed clean-ups.
How the connection works, in plain terms
About two minutes in Orbit, and you don’t touch any code.
- In the Orbit dashboard, go to Settings → AI Assistants.
- Copy the address shown there and paste it into your assistant’s Connectors setting (some apps call it Integrations or Apps).
- The assistant sends you to an Orbit approval screen. Tick the permissions it may have: reading products, orders and customers, and making the changes you allow. It gets nothing you did not tick.
- Use it. You can edit or revoke the connection at any time from the dashboard, and changes apply within seconds.
You get hundreds of actions across products, orders, customers, inventory, discounts, returns, shipping, pages and reports, on the same permissions as your own login. Detail on our AI page; the data rules are in the AI assistant data terms (version 2026-08-31).
Does it work with ChatGPT, Claude, or both?
Both, today. OpenAI’s developer documentation describes connecting a custom MCP server to ChatGPT: turn on developer mode under Settings, then add your server URL (developers.openai.com, checked 15 September 2026). Anthropic’s support article says custom connectors using remote MCP “are available on Claude, Cowork, and Claude Desktop for users on Free, Pro, Max, Team, and Enterprise plans”, free users limited to one, added under Customize → Connectors (support.claude.com, “Getting started with custom connectors using remote MCP”, checked 15 September 2026).
One caveat on cost, in our own dashboard’s words: “Adding your own connector is a paid feature in both Claude and ChatGPT.” Anthropic’s page today reads more generously for a single connector, and OpenAI’s consumer help pages returned a 403 when we tried to fetch them on 15 September 2026, so check your assistant’s help pages for custom connectors before you plan around a price. The Orbit side isn’t in question: connecting an assistant is included on every plan, not an add-on app. On Shopify, WooCommerce or anything else, ask the same questions: do you publish a connector, what can it change, can I revoke it?
Nebula: the other route, without leaving your shop
You don’t have to use an outside assistant at all. Nebula is Orbit’s assistant inside the dashboard, included on every plan. It adds products from a sentence, builds page sections, restyles the shop, answers data questions with a chart and a CSV, runs reports, writes product descriptions and blog posts, drafts discounts, and runs multi-step plans you approve before they execute. It runs on Claude Sonnet 5 by default, Claude Opus 5 selectable when a job deserves more care, with images from Google’s Gemini image model.
The published allowance is roughly 750 product descriptions a month on Solo (£25), 1,500 on Team (£85), 3,750 on Pro (£249) and 18,000 on Enterprise (£999), all ex VAT. The difference between the routes: Nebula needs no setup, while ChatGPT or Claude can hold your shop alongside everything else you’ve connected. Plenty of people use both.
What does this replace: your admin, your apps, or your VA?
It replaces admin tabs, and the mechanical half of what a virtual assistant does. It does not replace judgement.
Worked example, assumptions stated so you can argue with them. Say you have 40 products to add from a supplier spreadsheet, each with a title, description, price, stock figure and two images. By hand at a genuine 6 minutes per product, that’s 240 minutes, or 4 hours. At £18 an hour for a UK VA, £72. Ask an assistant with safe-write permission to create all 40 as drafts and it’s one instruction plus roughly 25 minutes of you reading the output and fixing what it got wrong: call it £7.50 of your own time at the same rate. The saving is real, £64.50 on this one job, but the 25 minutes did not disappear. That is the check, and it is the part people skip.
What it does not do: decide which 40 products are worth stocking, whether 15% off is the right discount, or what to say to an unhappy customer. None of that was ever the tab-clicking part of the job.
It earns its keep fastest on the catalogue work behind selling on more than one channel: rewriting titles to suit shopping feeds, filling the gaps that stop a product feed validating, getting enough attributes in to qualify for free Google product listings. Hours of typing with no thinking in it, which is exactly what you want to hand over. Our guide to multichannel selling in the UK covers the rest.
The honest limits
It can be wrong. Confidently. It will occasionally invent a product attribute, miscount a total, or misread which spreadsheet column held the price. Read the output on anything that matters, and start any new kind of job with a batch of ten rather than a thousand.
It acts on what it reads. People underestimate this one, so here it is in the words we use in the dashboard: “An assistant acts on what it reads. Product descriptions, order notes and customer messages are all text that can steer it, so give change permissions only to assistants you trust.” If a stranger can type text into your shop, a stranger can put instructions in front of your assistant. Anthropic makes the same point from the other direction, warning that malicious servers “may include hidden instructions that try to make Claude perform unintended actions” (support.claude.com, checked 15 September 2026). Read permission is close to harmless here. Live write is not.
Keep an escape route. You can export your products, orders and customers as CSV from Orbit at any time. Taking that export before a big bulk job costs nothing and occasionally saves your weekend.
The data rules are yours, not the assistant’s. Under our AI assistant data terms you stay the data controller, Orbit is your processor and the assistant provider becomes a further processor. Two lines are worth reading before you connect. On marketplaces: “You must not use an AI assistant to export, copy, keep or build reports and dashboards from marketplace order data or marketplace buyer data, unless that marketplace allows it.” On revocation: “You can revoke this connection at any time from your dashboard. Revoking stops all future access straight away. It cannot pull back anything the assistant has already received.” The terms also tell you to ask your provider whether they keep your prompts, for how long, and whether they train on them. Ask before you connect, not after.
Questions people ask about running a shop from ChatGPT or Claude
Can I really run my online shop from ChatGPT or Claude?
You can run the admin side of it. Connect your shop and you can ask questions about sales and stock, create and edit products, build pages, draft discounts and run reports by typing instead of clicking. You still approve anything that moves money or goes out to customers, and you still decide what to sell and at what price.
What is MCP in plain English?
MCP is an open standard that lets an AI application connect to an outside system, described on its own site as “an open-source standard for connecting AI applications to external systems” (modelcontextprotocol.io, checked 15 September 2026). Because it is a standard rather than a private integration, one connection published by your shop platform works with any assistant that supports it.
Is it safe to let an AI assistant change prices or refund orders?
Price changes are fine if you ask for a preview and approve the list first. Refunds we would not automate at all. An assistant acts on the text it reads, including text strangers can write into your shop, so anything irreversible should need a human yes. Orbit splits permissions into read, safe write and live write precisely so you can draw that line.
Does connecting my shop to ChatGPT or Claude cost extra?
Not on the Orbit side: it’s included on every plan, with no add-on app to buy. On the assistant side our dashboard notes that “adding your own connector is a paid feature in both Claude and ChatGPT”, though Anthropic’s support pages on 15 September 2026 listed custom connectors on Free through Enterprise plans with free users limited to one. Check your assistant’s help pages for custom connectors, as these rules change often.
Does this replace my virtual assistant?
It replaces the typing, not the thinking. Bulk product creation, description rewrites, feed clean-ups and report pulling all move to the assistant. Choosing products, setting discount strategy and handling an upset customer do not. Budget time for checking the work: on a 40-product import, expect roughly 25 minutes of review.
To try running a shop from ChatGPT or Claude without rebuilding anything: Orbit’s Settings → AI Assistants screen gives you the address, the permission tiers and a revoke button, and Nebula sits in the dashboard either way. Both are included on every plan, and the 14-day free trial needs no card, so you can wire it up, ask what sold last month, and decide from there: start free.